LogiScope-VQA - Operator Role: leaderboard
LogiScope-VQA Warehouse Knowledge Understanding task on operator roles (OR), asked over real logistics-park images and videos with human-validated answers (LogiScope-VQA, arXiv 2609.09790v1, Table 1). Micro-averaged accuracy converted from fractions to percent; higher is better; the paper's random-choice baseline scores 7.
Metric: Micro-averaged accuracy (%). Source: arxiv.org. Saturation forecast: Around September 2028. 20 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 Plus | 45 | #133 |
| 2 | Qwen 3.7 Plus | 40 | #98 |
| 3 | Gemini 3.1 Pro (Preview) | 30 | #64 |
| 4 | Claude Opus 4.7 | 30 | #51 |
| 5 | GPT-5.4 | 25 | #83 |
| 6 | Qwen 3 VL 235B A22B | 25 | #251 |
| 7 | Claude Sonnet 4.6 | 20 | #94 |
| 8 | GPT-5.5 | 20 | #33 |
| 9 | Kimi K2.6 | 20 | #113 |
| 10 | GLM-5.2 | 15 | #80 |
| 11 | GLM-4.7 | 15 | #202 |
| 12 | MiniMax-M2.5 | 15 | #316 |
| 13 | MiniMax-M2.1 | 15 | #299 |
| 14 | Kimi K2 (Thinking) | 5 | #252 (Kimi K2) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=logiscope-vqa-operator-role · How It Works · Data refreshed daily, snapshot 2026-11-05.